The use of epidemiologic data in the assessment of cancer risk.

Epidemiologic data can be extremely important in arriving at quantitative estimates of human cancer risk, expressed either as (a) attributable risk, or (b) relative risk. While case-control studies have a low cost/benefit ratio, cohort studies are more likely to yield data relating risk to amount of exposure, and to allow precise calculation of risk. However, cohort studies are long, expensive, and difficult to administer. Attrition and resultant unrepresentative composition can be major problems. Cohorts should be followed to extinction to permit accurate calculation of risk, which must take into account prolonged latency periods, competing risks, and the likelihood of increasing risk with time (or age). In view of certain limitations in epidemiologic and virologic methods, critical re-evaluation of the viral and oncogene theories are called for. The use of animal data for assessing human cancer risk requires further investigation of the assumptions and facts supporting two kinds of extrapolation: That from high to low dose carcinogen exposure, and cross-species extrapolation (i.e., from experimental animals to man). Large pieces of these two approaches rest on testable hypotheses which deserve further attention.